Compressor blades often have a small 'spike' in the surface pressure distribution at the leading edge. This may result from blade erosion, manufacture defects or compromises made in the original design process.

In this thesis it is shown that these spikes will increase the loss generated by a blade only when they become large enough to initiate boundary layer transition at the leading edge through a separation bubble; this process increases profile loss by about 30%. A criterion is presented, based on the spike diffusion, which can be used to determine whether leading edge transition will occur or not; this criterion is simple and quick to determine and has to potential to be used on a production line to reject those leading edges that would otherwise be detrimental.

The spikes are also shown to have a significant effect on the flow close to the endwalls. If they cause leading edge transition in this region then they will cause a growth in the size of the three-dimensional separations that exist in the corner between the blades' suction surfaces and the endwalls. On the low speed stator tested this process increased hub loss by around 100%.

Thus to prevent spikes becoming large a new method for producing a 'spikeless' leading edge has been developed; this leading edge can be attached easily to the thickness distribution of any blade and was found to be sharp, that is with very high curvature at the leading edge point.

This spikeless leading edge was also found to be the best when the effects of real manufacture deviations, measured off of a production line, were considered. Asymmetry was found to be detrimental and bluntness was only beneficial when unrealistically large deviations were considered. The best leading edge geometry is therefore sharp and symmetric.